1

enter image description hereThe figure below shows the PCA projections of inputs which are 14 meteorological features, (i.e. wind, temperature, humidity, pressure, and so on.) I would like to use any technique to make it more separable than this, The ISOMAP method is also used instead of PCA, but it gave a non-separable distribution as well! Any suggestions would be highly appreciated.

mhdella
  • 111
  • 2
  • 1
    have you considered trying to use an autoencoder? – spektr May 22 '16 at 14:51
  • Excuse my naivety...but what is an autoencoder? – mhdella May 22 '16 at 21:07
  • An autoencoder is a nonlinear approach to dimensionality reduction. It's an unsupervised learning technique that can, in its simplest form, just be a feed forward neural network formed a certain way. I would recommend googling it. – spektr May 22 '16 at 21:10
  • Thanks choward for your swift response and clarification.... I hope autoencoder would be better than what isomap was.... – mhdella May 22 '16 at 21:16
  • The point of these methods is to make data more separable, conditional on the fact they can be separated. I don't know much about ISOMAP, but I know an autoencoder can fair well for these types of tasks. Hopefully your data can actually be separated like you wish. – spektr May 22 '16 at 21:26
  • I really appreciate your clarification and suggestion, thanks choward – mhdella May 22 '16 at 21:37
  • Is this data set public so one could try to make it more separable? – spektr May 23 '16 at 16:09
  • No this data set is not public, but it's similar to any kind of other metrological data (wind speed, temperature, pressure, humidity, and so on) for the complete year with an hourly resolution. – mhdella May 24 '16 at 06:20

0 Answers0